Best LLMs for query_generation autogenerated — DTP Benchmark
Models
Frontier on this task: GPT-5.5 at 8.99 / 10. Quality bar at 95%: 8.54.
point-estimate floor (CI low) · upper CI (less certain) · Bars sorted by blended cost; cheapest qualifier first. Greyed rows are MEDIUM+ models whose point estimate clears the bar but whose CI low does not.
Cost breakdown
| Model | Quality | Sample | Blended cost / call | Savings vs best | Mode |
|---|---|---|---|---|---|
| Kimi K2.6 Moonshot AI | 8.59 / 10 CI [8.39, 8.78] | n=83 · ranked | $0.004194 | 81% cheaper | batch |
| Gemini 3.1 Pro Preview Gemini | 8.67 / 10 CI [8.50, 8.84] | n=71 · ranked | $0.009015 | 60% cheaper | batch |
| GPT-5.5 best OpenAI | 8.99 / 10 CI [8.65, 9.33] | n=71 · medium | $0.022538 | (anchor) | batch |
Typical call shape for this task: 3459 input tokens → 926 output tokens, EMA-tracked from production traffic. Blended cost = (in × in_price + out × out_price), rounded to 6 decimals.